Br. J. Cancer (1984), 50, 687-692

Endometrial cancer following treatment for breast cancer: A case-control study in Denmark M. Ewertzl, S.G. Machado2, J.D. Boice, Jr.2 & O.M. Jensen' 1The Danish Cancer Registry, Institute of Cancer Epidemiology under the Danish Cancer Society, Landskronagade 66, DK-2100 Copenhagen, Denmark; and 2Radiation Epidemiology Branch, National Cancer Institute, Bethesda, Maryland 20205, USA. Summary To evaluate the risk of endometrial cancer subsequent to breast cancer, a case-control study was carried out in Denmark. Between 1943-1977, 115 cases of histologically confirmed endometrial carcinoma developed more than 3 months after the diagnosis of a primary breast cancer in 51,638 women. A total of 235 breast cancer patients with no second primary cancer were matched to the cases on age, calendar year of diagnosis, and survival with an intact uterus. Identification of cases and controls relied upon records available in the Danish Cancer Registry. Information on risk factors and reproductive histories was abstracted from hospital records. Increased relative risks (RR) for endometrial cancer were associated with menopausal oestrogen use (RR=4.9), nulliparity (RR 2.1), late age at natural menopause (RR 2.9), and pelvic irradiation (RR= 1.4). No association was apparent for drugs used to treat breast cancer. This study indicates that breast and endometrial cancer share several common aetiologic factors and that studies of second primary cancers have the potential to provide information on risk factors other than those associated with therapy. =

Several studies indicate that breast cancer patients have an increased risk of developing subsequent endometrial cancer (Schoenberg, 1977; Kelsey & Hildreth, 1983). This may reflect a common aetiology of the two cancers (Armstrong, 1979) as well as the influence of various treatments for breast cancer (Hoover et al., 1976). Such therapies include irradiation to the pelvic area and sex hormones, especially oestrogen. The risk of endometrial cancer associated with oestrogen therapy for breast cancer was previously investigated in a cohort analysis by Hoover et al. (1976). They reported a 3- and 2-fold increase in risk, respectively, when oestrogens were given during an initial or subsequent course of breast cancer treatment. Their data on oestrogen use, however, derived from information in cancer registry records, and all treatments classified as "hormones" were presumed to be oestrogens. Moreover, no specification of type of oestrogen, dose and duration of treatment was available. A case-control study of endometrial cancer in women previously diagnosed with breast cancer, conducted simultaneously in Denmark and USA, was initiated to evaluate further the risk associated with oestrogen therapy and other factors suspected to affect the risk of endometrial cancer. This paper communicates the results of the Danish series. Correspondence: M. Ewertz Received 6 June 1984; accepted 6 August 1984.

=

Materials and methods

Both cases and controls were selected from the Danish Cancer Registry, which has a virtually complete registration of all cancers occurring in the entire Danish population since 1943. The case group was defined as women who developed a primary endometrial cancer at least 3 months after a diagnosis of breast cancer between 1943-1977. Among 51,638 women diagnosed with breast cancer, 138 cases of endometrial cancer were identified. Twelve cases (8.7%) were excluded because the source of exposure information (hospital records see below) had been destroyed or could not be located. Six women with sarcomas, 4 with other mixed tumours, and one with no histology report were excluded, leaving 115 cases with histologically confirmed carcinomas of the endometrium available for analysis. Among breast cancer patients with no subsequent malignancies, 235 were chosen as controls. These were individually matched to the cases on the following criteria: calendar year of breast cancer diagnosis (within same 5-year calendar period, e.g. 1943-47, 1948-52 etc.), age at breast cancer diagnosis (±3 years), and length of survival with an intact uterus, A control must have survived for at least as long as her corresponding case, and must not have had a hysterectomy before the date the case developed endometrial cancer. Information on treatment for breast cancer [surgery, radiation, chemotherapy, and hormones -

() The Macmillan Press Ltd.,

1984

688

M. EWERTZ et al.

(oestrogens, androgens, other steroids)], and potential risk factors for endometrial cancer (marital status, height, weight, diabetes, hypertension, parity, ages at menarche and menopause, use of oestrogens and oral contraceptives) was abstracted from the hospital records of cases and controls by one of the authors (ME). To verify the endometrial cancer diagnosis, pathology reports were examined and information on histology, grade and stage of disease obtained. Comparisons between cases and matched controls for all variables of interest were made by the conditional logistic regression methods described in Breslow & Day (1980). The computer program used (Lubin, 1981) provides estimates of relative risks (for both dichotomous and categorized variables) and corresponding estimates of precision. If the 95% confidence interval does not contain 1.0, then the relative risk is considered significant at the 5% level. In the analysis of a particular variable, the information for each set (a case and her corresponding controls) is used unless the value is missing for the case, or for all the controls in the

these categorized variables, trend tests were made by taking the midpoint of each category as the representative value or score. For open-ended categories, for which there is some arbitrariness of choice, the scores were chosen so that all scores were equidistant from another. For example, the variable menopausal age was grouped into 4 categories: 54, and the scores were taken as: 42, 47, 52, and 57, respectively. One-sided statistical tests were presented in most instances where the factor under study, e.g. menopausal age, has been previously found to be positively associated with the risk of endometrial cancer. Results

set.

The mean age at breast cancer diagnosis was 59.8 years for cases and 59.5 years for controls, and the mean survival from breast cancer diagnosis to development of endometrial cancer and corresponding year for controls was 10.8 and 12.1 years respectively.

Continuous variables were grouped into "categories", i.e. non-overlapping intervals. Calculations of relative risks between each category and a chosen reference category were made using dummy indicator variables for the categories. For

Table I summarizes the risk of developing endometrial carcinoma associated with selected personal characteristics and medical conditions. Women who had never given birth had twice the risk of parous women. When the relative risk was

Table I Relative risk of endometrial cancer associated with various personal characteristics and medical conditions among breast cancer patients No. of No. of No. of Relative risk strata exposed exposed matcheda cases controls (95% Cl)b

Factors

Categories

Marital status Age at menarche Age at natural menopause

Never/ever